In digital image-plane holographic microscopy (DIPHM), the precision and reliability of measurement are limited by noises introduced in recording process. Until now, the noise reduction research was focused on noise with high frequency. Here, a method was proposed to reduce noises not only with high frequency, but also with low frequency. By using the denoising method, the signal-to-noise ratio (SNR) of reconstructed phase was improved at the same time. The denoising method was realized by optimizing the digital image-plane microscopic hologram. The characteristics of the first intrinsic mode function (IMF1) in bi-dimensional empirical mode decomposition (BEMD) were in good accordance with that of gray value information of interference fringes in digital image-plane hologram. Hence, the digital image-plane hologram was decomposed by BEMD to extract the IMF1, utilizing the IMF1 as the optimized digital image-plane hologram. After that, the optimized digital image-plane hologram was reconstructed to retrieve phase to measure the surface profile of tested samples. The surface profile of a standard nano-step was measured by DIPHM with BEMD. By comparing and analyzing the frequency spectrum and measured height profile with and without BEMD, the optimizing method was proved to realize the correctness in surface profile measurement and noise suppression of reconstructed phase.